agent memory architecture

Sleep-time compute is a specific instantiation of how to structure persistent memory for agents—combining trajectory analysis with memory store updates.

Tutorial explicitly discusses 'different types of short-term and long-term memory and how to utilize them for agents to follow business rules'—this is a concrete instance of context engineering applie

Directly discusses where agent memory should live in system architecture—central to how context is managed across agent sessions.

Demonstrates a specific architectural choice (version control as backing store) for managing agent long-term memory.

Describes short-term and long-term memory modules as discrete components in step 4 of workflow, treating memory as separable from agent logic

The proposed 'personal knowledgebase + shareable memories' structure is a concrete instantiation of how agent context should be organized across persistence layers

CrewAI's agent memory system is a concrete implementation of multi-turn context preservation across task sequences